Compare / head-to-head
Step 3.5 FlashvsStep 5 Preview
Step 5 Preview leads 1 of 1 shared benchmark. Step 5 Preview has the larger context window (1M tokens).
Benchmarks from cited public sources; pricing from official pages; status from official provider feeds.
Shared benchmarks
0 – 1
Step 5 Preview leads
Cheaper per token
—
list price, input + output
Larger context
Step 5 Preview
1M tokens
Providers
StepFun
same provider
Step 3.5 Flash
StepFun · released 2026-02-12
text
- Context
- 256K
- Max out
- —
- Input /1M
- —
- Output /1M
- —
- Cached /1M
- —
- Scores
- 19 · 5 core
Step 5 Preview
StepFun · released 2026-09-20
textvisionvideo
Quality
Benchmark matrix
1 shared · 16 only Step 3.5 Flash · 4 only Step 5 Preview| Benchmark | Step 3.5 Flash | Step 5 Preview | Δ | Edge |
|---|---|---|---|---|
| Reported by both · 1 | ||||
| LMArena Elo | 1403.7 ↗ | 1453.7 ↗ | -50 | Step 5 Preview |
| Only Step 3.5 Flash reports · 16 | ||||
| SWE-bench Verified | 74.4% ↗ | not reported | — | — |
| AIME 2025 | 97.3% ↗ | not reported | — | — |
| AIME 2026 | 96.7% ↗matharena | not reported | — | — |
| BrowseComp | 51.6% ↗ | not reported | — | — |
| BrowseComp (with context manager) | 69% ↗ | not reported | — | — |
| BrowseComp-ZH | 66.9% ↗ | not reported | — | — |
| BrowseComp-ZH (with context manager) | 73.7% ↗ | not reported | — | — |
| GAIA (no file) | 84.5% ↗ | not reported | — | — |
| HMMT 2025 (Feb/Nov average) | 96.2% ↗ | not reported | — | — |
| IMOAnswerBench | 85.4% ↗ | not reported | — | — |
| LiveCodeBench v6 | 86.4% ↗ | not reported | — | — |
| ResearchRubrics | 65.3% ↗ | not reported | — | — |
| Tau2-Bench | 88.2% ↗ | not reported | — | — |
| Terminal-Bench 2.0 | 51% ↗ | not reported | — | — |
| xbench-DeepSearch 2025.05 | 83.7% ↗ | not reported | — | — |
| xbench-DeepSearch 2025.10 | 56.3% ↗ | not reported | — | — |
| Only Step 5 Preview reports · 4 | ||||
| LMArena Agent (LMArena) | not reported | 0.0048 ↗ | — | — |
| LMArena Vision (LMArena) | not reported | 1259.7 ↗ | — | — |
| LMArena WebDev (LMArena) | not reported | 1570.2 ↗ | — | — |
| Terminal-Bench 4.0 (Vals AI) | not reported | 31.8% ↗ | — | — |
Scores tagged "epoch" or "matharena" are independent runs, used only where the lab has not published its own; ⚠ marks rows MathArena flags as released after the competition. Higher is better on every row. Δ is Step 3.5 Flash minus Step 5 Preview in the benchmark's own unit. "Not reported" means the lab has not published that figure; it is not a zero. ↗ opens the source.
Specs & pricing
Side by side
Official pricing pages and model cards| Spec | Step 3.5 Flash | Step 5 Preview | Edge |
|---|---|---|---|
| Context window | 256K tokens | 1M tokens | Step 5 Preview |
| Max output | — | 64K tokens | — |
| Input price / 1M | — | $1 ↗ | — |
| Output price / 1M | — | $2.7 ↗ | — |
| Cached input / 1M | — | $0.05 ↗ | — |
| Input + output / 1M Lower is cheaper. List prices; batch, tool and regional fees excluded. | — | $3.7 | — |
| Modalities | text | text · vision · video | Step 5 Preview |
| Released | 2026-02-12 | 2026-09-20 | — |
| Cited benchmark scores | 19 | 5 | — |
Reliability
StepFun status
Live from /statusMore matchups
Step 3.5 Flash vs …
Models sharing the most benchmarksMore matchups
Step 5 Preview vs …
Models sharing the most benchmarksBuilt by Respan
Which one wins on your data?
Public benchmarks are a starting point. Run Step 3.5 Flash and Step 5 Preview on your own prompts with Respan evals, or route to either through one gateway key with automatic failover.